arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2302.02865cs.LGcs.AIstat.ML
收藏

概率对比学习恢复模糊输入的正确偶然不确定性

Probabilistic Contrastive Learning Recovers the Correct Aleatoric Uncertainty of Ambiguous Inputs

University of Tübingen(图宾根大学);TUM University(慕尼黑工业大学);Tübingen AI Center(图宾根人工智能中心)

Michael Kirchhof, Enkelejda Kasneci, Seong Joon Oh

AI总结:

针对模糊输入导致潜变量后验具有异方差不确定性的问题,扩展InfoNCE目标预测潜变量分布,证明其能恢复正确后验及偶然不确定性,并提供校准的不确定性估计与图像检索可信区间。

摘要:

对比训练的编码器最近被证明可以反演数据生成过程:它们将每个输入(例如图像)编码为生成该图像的真实潜变量向量(Zimmermann et al., 2021)。然而,现实世界的观测往往具有固有的模糊性。例如,图像可能模糊,或者只显示3D对象的2D视图,因此多个潜变量可能生成它们。这使得潜变量向量的真实后验具有概率性,并带有异方差不确定性。在这种设置下,我们扩展了常见的InfoNCE目标和编码器,以预测潜变量分布而不是点。我们证明,这些分布恢复了数据生成过程的正确后验,包括其偶然不确定性水平,直至潜空间的旋转。除了提供校准的不确定性估计外,这些后验还允许在图像检索中计算可信区间。它们包含与给定查询具有相同潜变量的图像,受限于其不确定性。代码可在https://github.com/mkirchhof/Probabilistic_Contrastive_Learning获取。

英文摘要:

Contrastively trained encoders have recently been proven to invert the data-generating process: they encode each input, e.g., an image, into the true latent vector that generated the image (Zimmermann et al., 2021). However, real-world observations often have inherent ambiguities. For instance, images may be blurred or only show a 2D view of a 3D object, so multiple latents could have generated them. This makes the true posterior for the latent vector probabilistic with heteroscedastic uncertainty. In this setup, we extend the common InfoNCE objective and encoders to predict latent distributions instead of points. We prove that these distributions recover the correct posteriors of the data-generating process, including its level of aleatoric uncertainty, up to a rotation of the latent space. In addition to providing calibrated uncertainty estimates, these posteriors allow the computation of credible intervals in image retrieval. They comprise images with the same latent as a given query, subject to its uncertainty. Code is available at https://github.com/mkirchhof/Probabilistic_Contrastive_Learning

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑